ChemRxiv
These are preliminary reports that have not been peer-reviewed. They should not be regarded as conclusive, guide clinical practice/health-related behavior, or be reported in news media as established information. For more information, please see our FAQs.
2019-03-24-Chem_RXiv-SubmitVersion.pdf (6.8 MB)
0/0

Hidden Bias in the DUD-E Dataset Leads to Misleading Performance of Deep Learning in Structure-Based Virtual Screening

preprint
submitted on 24.03.2019 and posted on 25.03.2019 by Lieyang Chen, Anthony Cruz, Steven Ramsey, Callum J. Dickson, José S. Duca, Viktor Hornak, David R. Koes, Tom Kurtzman

Recently much effort has been invested in using convolutional neural network (CNN) models trained on 3D structural images of protein-ligand complexes to distinguish binding from non-binding ligands for virtual screening. However, the dearth of reliable protein-ligand x-ray structures and binding affinity data has required the use of constructed datasets for the training and evaluation of CNN molecular recognition models. Here, we outline various sources of bias in one such widely-used dataset, the Directory of Useful Decoys: Enhanced (DUD-E). We have constructed and performed tests to investigate whether CNN models developed using DUD-E are properly learning the underlying physics of molecular recognition, as intended, or are instead learning biases inherent in the dataset itself. We find that superior enrichment efficiency in CNN models can be attributed to the analogue and decoy bias hidden in the DUD-E dataset rather than successful generalization of the pattern of protein-ligand interactions. Comparing additional deep learning models trained on PDBbind datasets, we found that their enrichment performances using DUD-E are not superior to the performance of the docking program AutoDock Vina. Together, these results suggest that biases that could be present in constructed datasets should be thoroughly evaluated before applying them to machine learning based methodology development.

Funding

R01-GM100946

R01-GM108340.

History

Email Address of Submitting Author

simpleliquid@gmail.com

Institution

Lehman College, City University of New York

Country

United States

ORCID For Submitting Author

0000-0003-0900-772X

Declaration of Conflict of Interest

Tom Kurtzman is founderand CSO of Deep Waters NYC, LLC.

Version Notes

Version 0001-A

Exports